Use inlined FNV1a hash for String
...and do not use __symbolHash(). Although currently the VM also
uses FNV1a hash for Symbols, the __symbolHash() does not handle
properly character with codepoint 0 (because '\0' is used as a
string terminator). This causes problems with Unicode16/32Strigs whose
version of FNV1a hash is using object size from header to determine
string's end.
Added Symbol>>hash that actually *uses* the __symbolHash() to make
sure it's hash is the the same as used bu the VM. Symbols with zeroes
are rare and there's no Unicode16/32Symbol.
This commit fixes issue #65.